RF Localization How good does it get?

1
RF LocalizationHow good does it get?

• Kamin Whitehouse
• UC Berkeley 2002

2
Is There a Better Way?

• You Bet!!

(But this is free)

• Infrared
• ultrasound
• TDOA

• Smallest
• Cheapest
• Easiest

3
Talk Outline
Regression
4
Best Fit
RSSI
50 cm
Distance
5
Reducing Noise
6
Reducing Error

• How does reducing noise effect error?

7
Results
With 200 samples, Std Error is 1cm
1000 samples/sec
200 samples in 0.2 sec
1cm/0.2 sec
5cm/sec
18 m/hr
8
Calibration
RSSI
Distance
9
Types of Node Variation
DC Bias
Radio/Ant Gain
Frequency
10
Auto Calibration
Can we learn the properties of nodes?
11
Auto Calibration
Maximize Consistency
12
Auto Calibration
Satisfy Constraints
13
Results
Manual calibration is accurate to 10cm.

• Only Accounts for DC Bias and Gain,
• Not Frequency

Auto calibration could do better.
14
Localization
Regression
Calibration
Localization
15
Geometric Approach
What if triangulation does not provide a unique
solution?
?
16
Statistical Approach
17
Future Work

• Deduce existence of attenuating obstacles
• Deduce existence of reflective surfaces
• Deduce orientation of antenna

18
RF LocalizationHow good does it get?

• Kamin Whitehouse
• UC Berkeley 2002